Clinical Commander

All dossiers
derm.cellulitis.core.v1

Cellulitis & erysipelas (dermatology lens)

dermatologyacutesubacuteadultacuteoutpatientinpatient

DERMATOLOGY-framed engine — intentionally distinct from id.cellulitis.core.v1 (ID empiric-spectrum lens). No duplicate: different engine_id, different primary question (skin-barrier + mimic + recurrence vs antibiotic spectrum + sepsis/necrotizing). Systemic sepsis, necrotizing fasciitis, and purulent/MRSA empiric-spectrum decisions are recognised then routed OUT by engine_id (id.sepsis.core.v1, id.necrotising-fasciitis.core.v1, id.cellulitis.core.v1) — not re-authored here. RxCUIs validated live against RxNav 2026-05-17: cephalexin 2231, dicloxacillin 3356, clindamycin 2582, penicillin V 7984, cefazolin 2180 (TMP-SMX 10831 / doxycycline 3640 / vancomycin 11124 / linezolid 190376 deferred to id.cellulitis.core.v1 spectrum engine — not used in this narrow-spectrum derm ladder). CREST 2005 cellulitis guideline is a document without a single stable PduMed PMID — cited via primary_guideline text; all 10 evidence.pmids are real, verifiable cellulitis/pseudocellulitis/prevention anchors. Bayesian linkage (pseudocellulitis pre-test priors, LR+/LR− for ≥8 distinguishing findings, conditional dependencies, T_treat/T_test, cross-dossier routing edges by engine_id) is documented in the co-located _design-brief.md + _research-bundle.md; first-class TS LR fields remain schema-blocked (same constraint as id.cellulitis.core.v1). Effect sizes (≥5): PATCH II HR 0.55 (95% CI 0.35-0.86); Webb compression HR 0.23 (95% CI 0.09-0.59); pseudocellulitis prevalence ~30% and 92% unnecessary-antibiotic rate (Raff JAMA Dermatol 2016); Hepburn 5 vs 10 d cure 98% vs 98% (non-inferior); Dupuy toe-web tinea OR ~2.4 for leg cellulitis; Levell pseudocellulitis ~33% of referrals.

Entry points (5)

  • symptom
    Unilateral warm, tender, expanding lower-leg erythema (IDSA 2014 SSTI Stevens — classic non-purulent cellulitis; NICE NG141 2019)
    unilateral_warm_erythema_lower_leg
  • symptom
    Sharply demarcated, raised, indurated erythema (erysipelas — superficial dermis + lymphatics; NICE NG141 2019; IDSA 2014 SSTI Stevens)
    sharply_demarcated_raised_facial_or_shin_erythema
  • symptom
    Acute "red leg" on a background of stasis dermatitis / lymphedema / leg ulcer / tinea pedis — high pseudocellulitis prior (Raff JAMA Dermatol 2016; David BJD 2011)
    red_leg_on_chronic_skin_disease
  • history
    ≥2 episodes/year cellulitis in the same limb — recurrent-cellulitis derm-prevention entry (Thomas NEJM 2013 PATCH II; NICE NG141 2019)
    recurrent_same_limb_cellulitis
  • problem_list
    Documented skin-barrier breach as portal of entry (interdigital tinea, toe-web maceration, eczema fissure, ulcer) (Dupuy BMJ 1999 — case-control of cellulitis risk factors)
    skin_barrier_breach_portal

Required inputs (16)

  • erythema_lateralityrequired
    symptom • used at ENTRY
    Bilateral lower-leg erythema is rarely infective cellulitis — strongest single pseudocellulitis discriminator (Raff JAMA Dermatol 2016 — bilaterality argues stasis dermatitis)
  • border_characterrequired
    symptom • used at CONTEXT
    Raised sharply demarcated border → erysipelas; indistinct → cellulitis; geometric/linear → contact dermatitis (NICE NG141 2019; IDSA 2014 SSTI Stevens)
  • pruritus_vs_painrequired
    symptom • used at CONTEXT
    Itch-dominant favors dermatitis (stasis/contact); pain/tenderness-dominant favors cellulitis (Raff JAMA Dermatol 2016 — pruritus has high LR for pseudocellulitis)
  • skin_portal_of_entryrequired
    history • used at CONTEXT
    Tinea pedis / interdigital maceration / eczema fissure / ulcer is the dominant modifiable derm risk factor and recurrence driver (Dupuy BMJ 1999; NICE NG141 2019)
  • chronic_edema_or_lymphedemarequired
    history • used at CONTEXT
    Lymphedema / chronic venous insufficiency is the strongest non-portal recurrence risk and a key mimic substrate (Cox BJD 2006; McNamara case-control)
  • venous_insufficiency_stasisrequired
    history • used at CONTEXT
    Stasis dermatitis is THE most common cellulitis mimic; chronic bilateral changes (hemosiderin, varicosities, lipodermatosclerosis) reframe the prior (Raff JAMA Dermatol 2016; Hirschmann JAAD 2012)
  • temperaturerequired
    vital • used at CONTEXT
    Fever raises infective prior and Eron class; afebrile bilateral itch-dominant red leg argues stasis dermatitis (Eron 2003; Raff JAMA Dermatol 2016)
  • necrotising_red_flagsrequired
    symptom • used at RED_FLAGS
    Pain out of proportion / dusky skin / bullae / crepitus / anesthesia → route OUT to id.necrotising-fasciitis.core.v1 (IDSA 2014 SSTI Stevens — surgical emergency, not authored here)
  • sbprequired
    vital • used at RED_FLAGS
    Hypotension / systemic toxicity → route OUT to id.sepsis.core.v1 (SSC 2021; Eron 2003 class IV — sepsis pathway not authored here)
  • diabetesrequired
    history • used at CONTEXT
    Diabetic skin barrier compromise + neuropathy alters portal, healing, and pathogen spectrum — derm wound-care threshold (IDSA 2014 SSTI Stevens; IWGDF 2023)
  • immunocompromise
    history • used at CONTEXT
    Atypical morphology + lower derm-referral and biopsy threshold (IDSA 2014 SSTI Stevens — broader pathogen diversity)
  • pregnancy
    history • used at TREATMENT
    Antibiotic-safety gating for the derm regimen (avoid doxycycline/TMP-SMX near term) (NICE NG141 2019)
  • wbc
    lab • used at INITIAL_WORKUP
    Normal WBC + bilateral itch-dominant erythema lowers infective prior (Raff JAMA Dermatol 2016 — inflammatory markers poorly discriminate but feed the chain)
  • crp
    lab • used at INITIAL_WORKUP
    Serial CRP trend supports response-to-therapy and IV→PO timing; very low CRP argues against cellulitis (NICE NG141 2019)
  • creatinine
    lab • used at TREATMENT
    Renal dosing for cephalexin / TMP-SMX in the derm PO regimen (IDSA 2014 SSTI Stevens)
  • compression_doppler_us
    imaging • used at BRANCHING_WORKUP
    Rule out DVT mimic when calf-dominant / pretest DVT non-trivial → route to cardio.dvt.core.v1 if positive (NICE NG141 2019)

12-phase flow (12)

  1. 1FRAME
    Frame as a SKIN-BARRIER infective process vs the dermatology mimic spectrum — ~30% of admitted "cellulitis" is pseudocellulitis (Raff JAMA Dermatol 2016; Levell BJD 2011). Erysipelas (superficial dermis + lymphatics, sharply raised border) vs cellulitis (deeper dermis/subcutis, indistinct border) (NICE NG141 2019). Systemic-sepsis and necrotizing concerns are routed OUT, not authored here.
    advance: derm scope confirmed; not-this-engine concerns routed by engine_id
  2. 2ENTRY
    Recognise acute unilateral warm erythema vs the recurrent-same-limb / red-leg-on-chronic-skin-disease entry; capture laterality up front (bilaterality is the single strongest pseudocellulitis flag) (Raff JAMA Dermatol 2016)
    inputs: erythema_laterality
    advance: entry trigger present; laterality recorded
  3. 3CONTEXT
    Skin-barrier substrate: tinea pedis / interdigital maceration / eczema fissure / leg ulcer (portal of entry — Dupuy BMJ 1999); chronic edema / lymphedema / venous-stasis dermatitis (mimic + recurrence substrate — Cox BJD 2006); border character; itch vs pain; fever; diabetes; immunocompromise. This phase builds the dermatology prior.
    inputs: border_character, pruritus_vs_pain, skin_portal_of_entry, chronic_edema_or_lymphedema, venous_insufficiency_stasis, temperature, diabetes, immunocompromise
    actions: workup.le_edema
    advance: derm substrate + pretest mimic prior assigned
  4. 4RED_FLAGS
    Necrotizing features (pain out of proportion, dusky/bullous skin, crepitus, anesthesia, rapid spread) → route OUT to id.necrotising-fasciitis.core.v1; systemic toxicity / hypotension / qSOFA≥2 → route OUT to id.sepsis.core.v1 with carryover state. These are recognised here but NOT managed here.
    inputs: necrotising_red_flags, sbp
    actions: calc.qsofa
    advance: necrotizing + sepsis red flags screened and routed by engine_id if positive
  5. 5INITIAL_WORKUP
    Mark and date the erythema border with a skin marker (the single most useful derm bedside tool — CREST 2005; NICE NG141 2019); photo-document; CBC + CRP baseline (poorly discriminating but anchor the response trend); creatinine for PO dosing. Cultures only if purulent/abscess/atypical (IDSA 2014 SSTI Stevens — blood cultures low-yield in uncomplicated cellulitis).
    inputs: wbc, crp, creatinine
    actions: panel.cbc, panel.inflammation, panel.renal
    advance: border traced + dated + photographed; baseline labs sent
  6. 6BRANCHING_WORKUP
    Pseudocellulitis decision tree: bilateral + itch + chronic stasis changes → stasis dermatitis (treat as dermatitis, NOT antibiotics); calf-dominant + pretest DVT → compression Doppler US, route to cardio.dvt.core.v1 if positive; acute woody tender plaque on chronic venous disease → acute lipodermatosclerosis; geometric/linear margin + exposure → contact dermatitis; monoarticular + crystals → gout. Skin biopsy / derm referral if diagnosis remains uncertain after 48-72 h or atypical.
    inputs: compression_doppler_us
    actions: workup.le_edema
    advance: mimic excluded or alternative dermatologic/vascular diagnosis assigned + routed
  7. 7DIFFERENTIAL
    Terminal derm differential with pivot findings: cellulitis vs stasis dermatitis (bilaterality + pruritus + chronicity pivot) vs acute lipodermatosclerosis (woody induration + medial-supramalleolar + no fever pivot) vs contact dermatitis (geometric margin + vesicles + itch pivot) vs DVT (calf tenderness + Wells + Doppler pivot — route to cardio.dvt.core.v1) vs erysipelas (raised sharply demarcated border + abrupt high fever pivot) vs lymphedema (chronic non-pitting + Stemmer pivot) vs gout (monoarticular + podagra + crystals pivot)
    advance: single best dermatologic diagnosis selected; co-existence (e.g., cellulitis ON stasis dermatitis) flagged
  8. 8RISK_STRATIFICATION
    Eron I-IV severity → derm disposition (Eron 2003; CREST 2005). Layer the derm modifiers: extensive skin-barrier breakdown, ulceration, lymphedematous limb, brittle diabetic foot skin, recurrent-same-limb pattern → lower outpatient threshold and trigger recurrence-prevention pathway.
    inputs: temperature, sbp
    actions: calc.qsofa
    advance: Eron class + derm modifier overlay assigned
  9. 9TREATMENT
    DERM-framed treatment: (1) confirm it IS cellulitis — do not antibiotic-treat stasis dermatitis (most common error; Raff JAMA Dermatol 2016 — 92% pseudocellulitis received unnecessary antibiotics); (2) narrow-spectrum streptococcal/MSSA-targeted PO for non-purulent (cephalexin / dicloxacillin; flucloxacillin UK) 5-7 d (Hepburn 2004 — 5 vs 10 d non-inferior; NICE NG141 2019); (3) SKIN-BARRIER repair in parallel — treat tinea pedis (topical/oral antifungal), emollient barrier restoration, limb elevation, compression once acute settles; (4) erysipelas → penicillin-class preferred (streptococcal); (5) purulent/MRSA-risk spectrum decisions routed to id.cellulitis.core.v1. Antibiotic-safety gating: pregnancy avoids doxycycline/TMP-SMX near term.
    inputs: creatinine, pregnancy, skin_portal_of_entry
    advance: diagnosis confirmed cellulitis; narrow-spectrum agent + skin-barrier plan started; 48-72 h derm recheck booked
  10. 10DISPOSITION
    Eron I → outpatient + 48-72 h derm/skin recheck with marked border; Eron II → outpatient with early review or ambulatory IV (OPAT); Eron III/IV or necrotizing/sepsis → admit and route OUT (id.cellulitis.core.v1 / id.necrotising-fasciitis.core.v1 / id.sepsis.core.v1). Pseudocellulitis confirmed → discharge OFF antibiotics with dermatitis plan.
    inputs: temperature, sbp
    advance: disposition documented; antibiotics stopped if pseudocellulitis
  11. 11MONITORING
    Border-trace progression at 48-72 h against the dated mark (CREST 2005); expect erythema to peak/worsen in first 24-48 h before improving even on correct therapy (NICE NG141 2019 — counsel to prevent premature escalation); CRP trend; if no improvement at 48-72 h reassess for mimic / abscess / wrong-spectrum / necrotizing before extending duration.
    inputs: crp, wbc
    actions: panel.inflammation
    advance: objective improvement by border + symptoms at 48-72 h, OR re-evaluation triggered
  12. 12FOLLOWUP
    DERMATOLOGY RECURRENCE PREVENTION (the core differentiator of this engine): eradicate + maintain tinea pedis control (recurrent risk if untreated — Dupuy BMJ 1999); lifelong emollient barrier repair for eczema/xerosis; compression + lymphedema therapy (decongestive) to reverse the recurrence substrate (Webb NEJM 2020 — compression halved recurrence); chronic venous insufficiency / stasis-dermatitis derm management; PATCH II penicillin V 250 mg BID × 12 mo if ≥2 episodes/year (Thomas NEJM 2013 — 45% relative recurrence reduction); derm referral for recurrent / atypical / ulcerated / diagnostically uncertain disease.
    inputs: skin_portal_of_entry, chronic_edema_or_lymphedema
    advance: recurrence-prevention + skin-barrier maintenance plan documented; derm referral made if criteria met